Norn Group

The next breakthroughs in biological AI will be shaped by the datasets we build now.

We’re planning an Impetus Grants @impetusgrants round to fund shared, AI-enabling biological datasets.

Learn more and contribute: norn.group/impetus-grants…

Agreed! We funded Columbia’s VIBRANT study via @ImpetusGrants because we wanted to move ovarian aging research into human trials.

Led by Yousin Suh (@YousinSuh) and Zev Williams, the trial is testing whether low-dose rapamycin can delay menopause by slowing ovarian aging, and what that could mean for women’s reproductive health.

More info here: clinicaltrials.gov/study/NCT05836…

Some cell populations become more common with age. Others become less common. Why is that? Can we use what changes in these populations to understand how tissues age?

We funded Junyue Cao @junyue_cao at The Rockefeller University via @impetusgrants to study how programmed cell death and cell populations change across cell types as mice age.

His team has given the field two large, publicly accessible datasets, presented in two Science papers.

In the first paper, Cao and colleagues measured gene activity in almost 22 million cell profiles across 14 mouse organs and tissues. They found that some cell populations became less common during early adulthood, while several immune cell populations became more common later in life.

They also looked at mice without functional B and T cells and compared them with normal mice. The effects differed between cell populations. Certain kidney and lung populations did not expand with age in these mice. Most populations that became less common with age followed the same pattern as in normal mice.

In their second paper, Cao and his colleagues measured how accessible different regions of DNA were to gene-regulating machinery in cells from 21 tissues. Some changes were shared across tissues, while others differed between sexes.

The relative abundance of one kidney cell type, for example, increased with age in female mice but decreased in males.

The DNA-accessibility atlas also revealed that areas associated with regulating inflammatory genes generally became more accessible with age in many cell types.

Read the papers:
doi.org/10.1126/scienc…
doi.org/10.1126/scienc…

Who tests the assumptions behind aging biomarkers? How do we know they measure aging the same way in different people?

We funded Alan Cohen via @impetusgrants to compare markers of immune aging across different populations.

Inflammaging is the increase in chronic inflammation that happens as we get older. Previous research in Italy had identified a pattern of inflammatory blood markers that was associated with aging and chronic disease. But how universal was that signature?

In Nature Aging, Alan Cohen, Michael Gurven and colleagues compared inflammatory markers in four populations. In Italy and Singapore, levels of inflammatory blood markers were higher among older adults and those with chronic diseases (especially kidney disease). In a Bolivian and a Malaysian population, however, inflammatory markers were not (or only weakly) related to age and the relationships with chronic diseases did not generalize.

Even Italy and Singapore differed on individual markers. IL-6, commonly used as a marker of inflammaging, was not associated with age in the Singaporean population.

Read the paper: doi.org/10.1038/s43587…

"Alphafold solved this subset of static structure prediction with a dataset that took 50 years to collect."

If we hope AI will help with aging, we should start generating the data it will need now. If we wait until we need them, we’ll be disappointed.

It’s not a question of whether we should pause AI development and instead just collect data. We can take a micro fraction of the overall budget and use it to generate the right data in parallel with everything else.

The next Impetus Grants @ImpetusGrants round we're setting up is a direct path to that. Our goal is $15M to get these datasets started, and we’ve raised ~$9M so far. norn.group/impetus-grants

We already anticipate compute demand and build data centers years before we need them. We should do the same for training and verification data on aging and disease.